Re: [Flexradio] frequency calibration etc
Hi Eric - To your earlier post regarding implementing within an FPGA the circuitry Alberto pointed to. Yes, this is easily done. In fact, I'd think you could improve upon the design, too. For example, in the schematic you really don't want the 43K resistor across the 470 uF cap - it, in series with the 15k resistor, will continually discharge the cap, meaning that the VCO control voltage (and thus frequency) will continually vary as the cap discharges and the phase-comparator pumps it back up to regain phase-lock. Ideally, if you're in lock, you would like the control voltage to be an unvarying DC level. Anyway - you could certainly implement all the digital circuitry as well as a '4046-style phase comparator within the fpga and drive an external loop filter, similar to shown in the schematic. Or...you could even attempt loop filtering within the fpga and generate the VCO control voltage a number of ways - drive a dac, for example (similar to Shera's design - which I use here to drive an HP 106B, by the way). But no matter which route is followed, much attention needs to be paid to ground power routing, layout, etc, to ensure that minimal noise is added to the VCO control voltage from external sources. - Jeff, WA6AHL
Re: [Flexradio] frequency calibration etc
At 05:14 AM 11/23/2005, Jeff Anderson wrote: Hi Eric - To your earlier post regarding implementing within an FPGA the circuitry Alberto pointed to. Yes, this is easily done. In fact, I'd think you could improve upon the design, too. For example, in the schematic you really don't want the 43K resistor across the 470 uF cap - it, in series with the 15k resistor, will continually discharge the cap, meaning that the VCO control voltage (and thus frequency) will continually vary as the cap discharges and the phase-comparator pumps it back up to regain phase-lock. Ideally, if you're in lock, you would like the control voltage to be an unvarying DC level. Turning the first order loop into a second order loop. A first order loop will always have some small phase error, but it will be reasonably constant (frequency dependent, possibly). Anyway - you could certainly implement all the digital circuitry as well as a '4046-style phase comparator within the fpga and drive an external loop filter, similar to shown in the schematic. Or, use a fixed oscillator, and run an NCO in the FPGA to create an offset frequency, which you then mix with the fixed oscillator to create your locked output. [This is what we are doing in an experimental deep space transponder.. where phase noise is of obssessive concern] Or...you could even attempt loop filtering within the fpga and generate the VCO control voltage a number of ways - drive a dac, for example (similar to Shera's design - which I use here to drive an HP 106B, by the way). But no matter which route is followed, much attention needs to be paid to ground power routing, layout, etc, to ensure that minimal noise is added to the VCO control voltage from external sources. Which is precisly why I like the idea of measuring the offset and compensating in other ways, rather than steering the oscillator itself. Then, you can work on getting the best possible performance from the oscillator, which can be highly isolated from the outside world. - Jeff, WA6AHL Jim, W6RMK
Re: [Flexradio] frequency calibration etc
Somewhere, on one of the partitions on one of the drives visible to one of the OSs on one of my computers, I have a complete NCO written in VHDL pretty much ready for dropping into an FPGA (you do the grunge work of assigning ports, etc.). IF I remember correctly (guaranty does not cover memory), an Altera AE found it for me. If anyone wants to try it contact me directly and I will look for it. Phil, K3IB - Original Message - From: Jim Lux [EMAIL PROTECTED] To: Jeff Anderson [EMAIL PROTECTED]; [EMAIL PROTECTED] Biz flexradio@flex-radio.biz Sent: Wednesday, November 23, 2005 10:10 AM Subject: Re: [Flexradio] frequency calibration etc Or, use a fixed oscillator, and run an NCO in the FPGA to create an offset frequency, which you then mix with the fixed oscillator to create your locked output. [This is what we are doing in an experimental deep space transponder.. where phase noise is of obssessive concern]
Re: [Flexradio] frequency calibration etc
Hi Jim,Jim Lux [EMAIL PROTECTED] wrote: Or, use a fixed oscillator, and run an NCO in the FPGA to create an offset frequency, which you then mix with the fixed oscillator to create your locked output. [This is what we are doing in an experimental deep space transponder.. where phase noise is of obssessive concern] [WA6AHL] :I likeyour idea. But let me see if I understand it...inan app such as, say,a general-purpose way of generating a stable frequency using the Jupiter 1pps as reference, are you saying that the NCO (with an external, stable, oscillator as its clock source) would, in essence,be the digital version of the preiously mentioned VCO? Phase comparison between the NCO's output and the reference1ppsis done within the FPGA and the error used to "steer" the NCO and proved an output that's locked to the ref? Is one ofthe tradeoffs low phase-noise vs. frequency-step "quantization" of the NCO? (E.g. the NCO mightnever be *exactly* on frequency). In an application specific to the SDR1K, per Bob's example, youdon't need the NCO. Instead, feed the error sig back to the SDR1K and let s/w handle frequency correction... Or...you could even attempt loopfiltering within the fpga and generate the VCO control voltage a number ofways - drive a dac, for example (similar to Shera's design - which I usehere to drive an HP 106B, by the way). But no matter which route isfollowed, much attention needs to be paid to ground power routing, layout,etc,to ensure that minimal noise is added to the VCO control voltage fromexternal sources.Which is precisly why I like the idea of measuring the offset and compensating in other ways, rather than steering the oscillator itself. Then, you can work on getting the best possible performance from the oscillator, which can be highly isolated from the outside world. [WA6AHL] Agreed. Of course, depending upon how sensitive to noise your application is, good layout bypassing techniques still apply even for theNCO technique. Given finite slew-rates of digital signals, ground bounce or supply sag can increase switching-threshold uncertainty, resulting in jitter in the digital domain.Jim Lux [EMAIL PROTECTED] wrote: At 05:14 AM 11/23/2005, Jeff Anderson wrote:Hi Eric -To your earlier post regarding implementing within an FPGA the circuitryAlberto pointed to. Yes, this is easily done. In fact, I'd think you couldimprove upon the design, too. For example, in the schematic you reallydon't want the 43K resistor across the 470 uF cap - it, in series with the15k resistor, will continually discharge the cap, meaning that the VCOcontrol voltage (and thus frequency) will continually vary as the capdischarges and the phase-comparator pumps it back up to regain phase-lock.Ideally, if you're in lock, you would like the control voltage to be anunvarying DC level.Turning the first order loop into a second order loop. A first order loop will always have some small phase error, but it will be reasonably constant (frequency dependent, possibly).Anyway - you could certainly implement all the digital circuitry as well asa '4046-style phase comparator within the fpga and drive an external loopfilter, similar to shown in the schematic.Or, use a fixed oscillator, and run an NCO in the FPGA to create an offset frequency, which you then mix with the fixed oscillator to create your locked output. [This is what we are doing in an experimental deep space transponder.. where phase noise is of obssessive concern] Or...you could even attempt loopfiltering within the fpga and generate the VCO control voltage a number ofways - drive a dac, for example (similar to Shera's design - which I usehere to drive an HP 106B, by the way). But no matter which route isfollowed, much attention needs to be paid to ground power routing, layout,etc,to ensure that minimal noise is added to the VCO control voltage fromexternal sources.Which is precisly why I like the idea of measuring the offset and compensating in other ways, rather than steering the oscillator itself. Then, you can work on getting the best possible performance from the oscillator, which can be highly isolated from the outside world.- Jeff, WA6AHLJim, W6RMK
Re: [Flexradio] frequency calibration etc
Actually, I would be much more interested in the shell script you use to search all of the machines running different OS's to find the file ;-). I don't know what an AE is unless it means application engineer but be careful about IP ;-) intellectual property issues with these cores. With that taken care of, send it over. I always like looking at how others do their cordic arithmetic in these cores. Bob Philip M. Lanese wrote: Somewhere, on one of the partitions on one of the drives visible to one of the OSs on one of my computers, I have a complete NCO written in VHDL pretty much ready for dropping into an FPGA (you do the grunge work of assigning ports, etc.). IF I remember correctly (guaranty does not cover memory), an Altera AE found it for me. If anyone wants to try it contact me directly and I will look for it. Phil, K3IB - Original Message - From: Jim Lux [EMAIL PROTECTED] To: Jeff Anderson [EMAIL PROTECTED]; [EMAIL PROTECTED] Biz flexradio@flex-radio.biz Sent: Wednesday, November 23, 2005 10:10 AM Subject: Re: [Flexradio] frequency calibration etc Or, use a fixed oscillator, and run an NCO in the FPGA to create an offset frequency, which you then mix with the fixed oscillator to create your locked output. [This is what we are doing in an experimental deep space transponder.. where phase noise is of obssessive concern] ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz -- Laziness is the number one inspiration for ingenuity. Guilty as charged!
Re: [Flexradio] frequency calibration etc
I like your idea of keeping the existing low phase noise Oscillator and measure the drift and correct in the software. The whole thing could be quite cheap yet give you good results. Like you mentioned using a small CPU running off a GPS calibrated clock to measure the Oscillator output, multiple readings and average it out. This entails a minimum change to the radio, the software changes would be very small, adding an offset to the frequency of the DDS, a little bit of code to read the calibration offset from the measuring CPU. Basically a Huff and Puff using software to do the actual correction. At 11:04 AM 11/23/2005, you wrote: Hi Jim, Which is precisly why I like the idea of measuring the offset and compensating in other ways, rather than steering the oscillator itself. Then, you can work on getting the best possible performance from the oscillator, which can be highly isolated from the outside world. - Jeff, WA6AHL Jim, W6RMK ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz Cecil Bayona KD5NWA www.qrpradio.com I fail to see why doing the same thing over and over and getting the same results every time is insanity: I've almost proved it isn't; only a few more tests now and I'm sure results will differ this time ...
Re: [Flexradio] frequency calibration etc
At 09:04 AM 11/23/2005, Jeff Anderson wrote: Hi Jim, Jim Lux [EMAIL PROTECTED] wrote: Or, use a fixed oscillator, and run an NCO in the FPGA to create an offset frequency, which you then mix with the fixed oscillator to create your locked output. [This is what we are doing in an experimental deep space transponder.. where phase noise is of obssessive concern] [WA6AHL] : I like your idea. But let me see if I understand it...in an app such as, say, a general-purpose way of generating a stable frequency using the Jupiter 1pps as reference, are you saying that the NCO (with an external, stable, oscillator as its clock source) would, in essence, be the digital version of the preiously mentioned VCO? Phase comparison between the NCO's output and the reference 1pps is done within the FPGA and the error used to steer the NCO and proved an output that's locked to the ref? Phase comparison between the external clock and the 1pps is used to steer the NCO, but yes, the NCO replaces the VCO. Is one of the tradeoffs low phase-noise vs. frequency-step quantization of the NCO? (E.g. the NCO might never be *exactly* on frequency). Sure.. And there's also the issue of NCO spurs.. Nothing comes for free. In the space application, the nice thing is that it makes it easy to generate one frequency that is coherently locked to another reference. Here's the typical scenario. The deep space network sends a carrier at, say, 7160.000771 MHz to the spacecraft where it is used as a reference to lock an oscillator on the spacecraft. That oscillator is then used to synthesize a return carrier at, say, 8412.283950 MHz, which the ground receives. The ratio between frequencies is something like 880/749, called the turnaround ratio. The ground receiver tracking loop bandwidth might be one Hz or so (hey, we're talking about a 10 Watt transmitter from Pluto or something... you need all the help you can get). If the transponder on the spacecraft is good enough, you can do useful radio science by measuring the changes in the phase of the received signal. For instance, you can do things like measure the density of a planet or moon's atmosphere. Or, because measuring fractions of a cycle in phase is like measuring displacements on the order of centimeters, you can do orbit determination for things a long, long ways away. Historically, all this tracking and synthesizing was done in conventional analog PLL kinds of ways, with a VCXO and chains of multipliers. Naturally, because good, quiet VCXOs have small tuning ranges, you'd have to decide on your frequencies a long time in advance, and get a crystal ground for that, etc. Now, however, you could use a very stable and quiet XO, and track out the difference of the uplink signal against the XO entirely with a digital loop, and synthesize the coherent downlink also with a digital oscillator. The frequency of the underlying XO isn't as important, so you can use the same XO for lots of different missions (or, even, change the channel assignment late in the game). A very quiet XO is useful too, because sometimes, you're not doing coherent turnaround, but just generating the downlink from the internal oscillator. You'd like that downlink signal to be very quiet (so that you can do ranging, for instance, or because you're sending data at 8 bits/second). You'd also like the ability to control the frequency of that downlink without having to physically change the crystal frequency, and an NCO can do that. http://tmo.jpl.nasa.gov/progress_report/42-156/156C.pdf talks a bit about this kind of stuff. In an application specific to the SDR1K, per Bob's example, you don't need the NCO. Instead, feed the error sig back to the SDR1K and let s/w handle frequency correction... Exactly.. James Lux, P.E. Spacecraft Radio Frequency Subsystems Group Flight Communications Systems Section Jet Propulsion Laboratory, Mail Stop 161-213 4800 Oak Grove Drive Pasadena CA 91109 tel: (818)354-2075 fax: (818)393-6875
Re: [Flexradio] frequency calibration etc
Folks I have a stupid question which I should be able to look up. Can an FPGA pin actually accept a 10 mhz or 200 mhz signal so that the LEs could be configured to divide it down? I really do like Bobs example and suggestion. Have 1 10 mhz tcvcxo interfaced to the GPS and stabilized. Divide the 200 mhz signal down to say 10 meg compare the reference sig to the LO and tell the software to correct for variance in the 200 mhz LO. Am I understanding this correctly. (forget whether it is a PIC or FPGA or discrete hardware). I need a block diag to follow all this (smile). Its fun tho! Eric From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] On Behalf Of Jeff Anderson Sent: Wednesday, November 23, 2005 12:04 PM To: Jim Lux; [EMAIL PROTECTED] Biz Subject: Re: [Flexradio] frequency calibration etc Hi Jim, Jim Lux [EMAIL PROTECTED] wrote: Or, use a fixed oscillator, and run an NCO in the FPGA to create an offset frequency, which you then mix with the fixed oscillator to create your locked output. [This is what we are doing in an experimental deep space transponder.. where phase noise is of obssessive concern] [WA6AHL] :I likeyour idea. But let me see if I understand it...inan app such as, say,a general-purpose way of generating a stable frequency using the Jupiter 1pps as reference, are you saying that the NCO (with an external, stable, oscillator as its clock source) would, in essence,be the digital version of the preiously mentioned VCO? Phase comparison between the NCO's output and the reference1ppsis done within the FPGA and the error used to steer the NCO and proved an output that's locked to the ref? Is one ofthe tradeoffs low phase-noise vs. frequency-step quantization of the NCO? (E.g. the NCO mightnever be *exactly* on frequency). In an application specific to the SDR1K, per Bob's example, youdon't need the NCO. Instead, feed the error sig back to the SDR1K and let s/w handle frequency correction... Or...you could even attempt loop filtering within the fpga and generate the VCO control voltage a number of ways - drive a dac, for example (similar to Shera's design - which I use here to drive an HP 106B, by the way). But no matter which route is followed, much attention needs to be paid to ground power routing, layout, etc, to ensure that minimal noise is added to the VCO control voltage from external sources. Which is precisly why I like the idea of measuring the offset and compensating in other ways, rather than steering the oscillator itself. Then, you can work on getting the best possible performance from the oscillator, which can be highly isolated from the outside world. [WA6AHL] Agreed. Of course, depending upon how sensitive to noise your application is, good layout bypassing techniques still apply even for theNCO technique. Given finite slew-rates of digital signals, ground bounce or supply sag can increase switching-threshold uncertainty, resulting in jitter in the digital domain. Jim Lux [EMAIL PROTECTED] wrote: At 05:14 AM 11/23/2005, Jeff Anderson wrote: Hi Eric - To your earlier post regarding implementing within an FPGA the circuitry Alberto pointed to. Yes, this is easily done. In fact, I'd think you could improve upon the design, too. For example, in the schematic you really don't want the 43K resistor across the 470 uF cap - it, in series with the 15k resistor, will continually discharge the cap, meaning that the VCO control voltage (and thus frequency) will continually vary as the cap discharges and the phase-comparator pumps it back up to regain phase-lock. Ideally, if you're in lock, you would like the control voltage to be an unvarying DC level. Turning the first order loop into a second order loop. A first order loop will always have some small phase error, but it will be reasonably constant (frequency dependent, possibly). Anyway - you could certainly implement all the digital circuitry as well as a '4046-style phase comparator within the fpga and drive an external loop filter, similar to shown in the schematic. Or, use a fixed oscillator, and run an NCO in the FPGA to create an offset frequency, which you then mix with the fixed oscillator to create your locked output. [This is what we are doing in an experimental deep space transponder.. where phase noise is of obssessive concern] Or...you could even attempt loop filtering within the fpga and generate the VCO control voltage a number of ways - drive a dac, for example (similar to Shera's design - which I use here to drive an HP 106B, by the way). But no matter which route is followed, much attention needs to be paid to ground power routing, layout, etc, to ensure that minimal noise is added to the VCO control voltage from external sources. Which is precisly why I like the idea of measuring the offset and compensating in other ways, rather than steering the oscillator itself. Then, you can work on getting the best possible performance
Re: [Flexradio] frequency calibration etc
Can an FPGA pin actually accept a 10 mhz or 200 mhz signal so that the LE’s could be configured to divide it down? 10 MHz is no problem. 200 MHz! Many FPGAs can handle this frequency, some go faster, many can't go quite this fast. 73, Lyle KK7P
Re: [Flexradio] frequency calibration etc
Lyle Thanks. Well not too difficult to divide the 200 by 10 externally and perhaps provide buffering for the 200 mhz sig coming out of the SDR. Eric2 -Original Message- From: Lyle Johnson [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 23, 2005 5:11 PM To: ecellison Cc: 'Jeff Anderson'; 'Jim Lux'; '[EMAIL PROTECTED] Biz' Subject: Re: [Flexradio] frequency calibration etc Can an FPGA pin actually accept a 10 mhz or 200 mhz signal so that the LE's could be configured to divide it down? 10 MHz is no problem. 200 MHz! Many FPGAs can handle this frequency, some go faster, many can't go quite this fast. 73, Lyle KK7P
Re: [Flexradio] frequency calibration etc
Hello Eric! Thanks. Well not too difficult to divide the 200 by 10 externally and perhaps provide buffering for the 200 mhz sig coming out of the SDR. The EP1C3T100 used in the Xylo board you mentioned earlier is good to 275/320/405 MHz depending on the speed grade of the part. That may be based on an internal PLL rather than an externally applied clock to a pin, I haven't dug that deeply into the data sheet. But the internal logic should have no problem at 200 MHz with careful design. 73, Lyle KK7P
Re: [Flexradio] frequency calibration etc
At 01:56 PM 11/23/2005, you wrote: Folks I have a stupid question which I should be able to look up. Can an FPGA pin actually accept a 10 mhz or 200 mhz signal so that the LEs could be configured to divide it down? Yes.. depends on the FPGA, though, what the maximum clock rate is. Jim 'rmk
Re: [Flexradio] frequency calibration etc
At 02:15 PM 11/23/2005, ecellison wrote: Lyle Thanks. Well not too difficult to divide the 200 by 10 externally and perhaps provide buffering for the 200 mhz sig coming out of the SDR. Eric2 Even better, if you don't want to give up the resolution (dividing by 10 does that) is divide by 10 and run that in one pin. Divide by 11 and run that in another pin. Now you can actually get the full resolution. Jim, 'rmk
Re: [Flexradio] frequency calibration etc
Jim Could we just not divide at all using the 1 pps gate and just use the variance of the last 2 or three digits, and make the assumption that it was accurate 200,000,xxx? I wouldn't think we would want to make any correction in software more frequently than one second. Or perhaps we could store a running average of the last three digits in an accumulator for the comparison against the 'rock'. Eric2 -Original Message- From: Jim Lux [mailto:[EMAIL PROTECTED] Sent: Wednesday, November 23, 2005 5:44 PM To: ecellison Cc: '[EMAIL PROTECTED] Biz' Subject: RE: [Flexradio] frequency calibration etc At 02:15 PM 11/23/2005, ecellison wrote: Lyle Thanks. Well not too difficult to divide the 200 by 10 externally and perhaps provide buffering for the 200 mhz sig coming out of the SDR. Eric2 Even better, if you don't want to give up the resolution (dividing by 10 does that) is divide by 10 and run that in one pin. Divide by 11 and run that in another pin. Now you can actually get the full resolution. Jim, 'rmk
Re: [Flexradio] frequency calibration etc
Relatively prime numbers are your friend. Jim Lux wrote: At 02:15 PM 11/23/2005, ecellison wrote: Lyle Thanks. Well not too difficult to divide the 200 by 10 externally and perhaps provide buffering for the 200 mhz sig coming out of the SDR. Eric2 Even better, if you don't want to give up the resolution (dividing by 10 does that) is divide by 10 and run that in one pin. Divide by 11 and run that in another pin. Now you can actually get the full resolution. Jim, 'rmk ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz -- Laziness is the number one inspiration for ingenuity. Guilty as charged!
Re: [Flexradio] frequency calibration etc
At 03:52 PM 11/23/2005, ecellison wrote: Jim Could we just not divide at all using the 1 pps gate and just use the variance of the last 2 or three digits, and make the assumption that it was accurate 200,000,xxx? I wouldn't think we would want to make any correction in software more frequently than one second. Or perhaps we could store a running average of the last three digits in an accumulator for the comparison against the 'rock'. Sure.. but the LSBs (which are the ones of interest) are the ones that have to count the fastest, so you still need to have something running at 200 MHz. Jim..
Re: [Flexradio] frequency calibration etc
At 04:09 PM 11/23/2005, Robert McGwier wrote: Relatively prime numbers are your friend. Indeed, but I was thinking that one might be able to adopt an off the shelf dual modulus counter, which typically differ by one count. Actually, there's all kinds of other strategies, too.. you can use the 200 MHz to clock a variety of counters, and pull out various lower rate clocks. Or, a fast counter that has a latch. Ultimately, I guess the real limit is in the setup/hold times in response to the 1pps (if that's what you're using) Jim Lux wrote: At 02:15 PM 11/23/2005, ecellison wrote: Lyle Thanks. Well not too difficult to divide the 200 by 10 externally and perhaps provide buffering for the 200 mhz sig coming out of the SDR. Eric2 Even better, if you don't want to give up the resolution (dividing by 10 does that) is divide by 10 and run that in one pin. Divide by 11 and run that in another pin. Now you can actually get the full resolution. Jim, 'rmk ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz -- Laziness is the number one inspiration for ingenuity. Guilty as charged! James Lux, P.E. Spacecraft Radio Frequency Subsystems Group Flight Communications Systems Section Jet Propulsion Laboratory, Mail Stop 161-213 4800 Oak Grove Drive Pasadena CA 91109 tel: (818)354-2075 fax: (818)393-6875
[Flexradio] Frequency calibration etc
Eric Can you please put me down for a Reflock and other items as well. Thanks Ross ZL1WN
Re: [Flexradio] Frequency calibration etc
Ross Well, I think we are a ways away from a kit. I was sort of suggesting that we try just using the Rockwell Jupiter board and its 10 mhz oscillator naked and just see what we get. In 2 lengthy arounds on this thread mostly what we have is theory and nothing tried at this point. I think that John N8UR is the only one I have heard about who is using an external 10 mhz reference to the SDR. Comment John? Anyone else using Geralds mod kit with external reference? Results? In the end, it looks like TAPR will have all the items we need for the reference except power and GPS. If the sound card is a big player in the variance as suggested, probably Phil1s experiments with the Altera and Wolfson chips with a NCO is the way to go, that way we could have accurate control of both the LO and Sound Card oscillators. Course I have not even started getting my mind around that project. At this point, I am so confused, with all the discussion, I dont know which way is up, and also which way to go! (smile). Somebody just needs to try the simple 10 mhz mod and see how it plays and where we need to go. I think the 10 mhz (I think sine wave) osc on the Rockwell board would be worth a shot. Eric From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]] On Behalf Of Ross Sent: Tuesday, November 22, 2005 1:00 AM To: FlexRadio@flex-radio.biz Subject: [Flexradio] Frequency calibration etc Eric Can you please put me down for a Reflock and other items as well. Thanks Ross ZL1WN
Re: [Flexradio] Frequency calibration etc
ecellison wrote: Ross Well, I think we are a ways away from a kit. I was sort of suggesting that we try just using the Rockwell - Jupiter board and it's 10 mhz oscillator 'naked' and just see what we get. In 2 lengthy arounds on this thread mostly what we have is theory and nothing tried at this point. I think that John - N8UR is the only one I have heard about who is using an external 10 mhz reference to the SDR. Comment John? Anyone else using Geralds mod kit with external reference? Results? Yes, I've been using my SDR-1000 with external reference from an HP 5065A Rubidium frequency standard. I haven't noticed any horrid phase noise problems from the 10MHz to 200MHz multiplication, but I also haven't been looking very hard for them. I hope that fairly soon -- perhaps over the Xmas holidays -- I will get temporary access to HP's latest super-duper phase noise measurement box, and one of the tests will be to compare the SDR-1000 with original crystal versus the external reference. I'll certainly post those results when I have them. 73, John
Re: [Flexradio] Frequency calibration etc
I too have been using an external 10 MHz source on my SDR. First I used a Rubidium source off of ebay, never noticed anything different from the internal crystal other than no drift. (rock solid) Note that I have only used it on 902 thru 24 GHz - where the SDR is the 28 MHz IF. I recently switched the 10 MHz source to the HP Z3801 GPS receiver. I A/B switched the Rubidium and Z3801A back and forth and could tell no difference in the audio. However when I used the waterfall display while doing the switching, the Rubidium source show a bit of light white noise on the screen (where I adjusted it) as the waterfall was moving (it was very slight). The 10 MHz Z3801A source seemed to eliminate this. I am not experiencing anything drastic that would indicate the improper reception or weak signals covered up while using these 10 MHz sources, but I have no way to measure phase noise, just a gut feel for how my system is operating on the UHF/SHF bands. 73 Mike - KM0T www.km0t.com Comment John? Anyone else using Geralds mod kit with external reference? Results?
Re: [Flexradio] Frequency calibration etc
You made the right decision to move from the rubidium. The off ebay rubidium sources almost all FM the line looking for it. This means that the output of one of these rubidiums is typically a zero mean on the frequency of interest and should never be used directly but only as a reference in taming oscillators and NEVER as the oscillator itself. Many of the more modern rubidium sources ($$$) do not show this effect. Tom sold me his very high stability Efratom FRK rubidium and it is just great FOR LOCKING FREQUENCY. I use it as the reference for my reflock I controlled oscillators since it stays stable when mobile whereas the GPS timing oscillators have the nice trick of going into survey mode if you move them. You should hear the 24 GHz system with that running! The goal was to tame the VCXO oscillators for the 10 and 24 GHz systems. Believe me, I have made all the dumb mistakes a person could make doing this. Thanks to Tom, I have my mind clear at last (well, on this topic anyway). The phase noise of the crystal oscillator part of the system provides phase stability. The rubidium part of the system provides ACCURACY not phase stability. Hey, out there in Iowa you don't have enough strong signal neighbors probably to even know you have a phase noise problem! It is there. Bob Mike King - KM0T wrote: I too have been using an external 10 MHz source on my SDR. First I used a Rubidium source off of ebay, never noticed anything different from the internal crystal other than no drift. (rock solid) Note that I have only used it on 902 thru 24 GHz - where the SDR is the 28 MHz IF. I recently switched the 10 MHz source to the HP Z3801 GPS receiver. I A/B switched the Rubidium and Z3801A back and forth and could tell no difference in the audio. However when I used the waterfall display while doing the switching, the Rubidium source show a bit of light white noise on the screen (where I adjusted it) as the waterfall was moving (it was very slight). The 10 MHz Z3801A source seemed to eliminate this. I am not experiencing anything drastic that would indicate the improper reception or weak signals covered up while using these 10 MHz sources, but I have no way to measure phase noise, just a gut feel for how my system is operating on the UHF/SHF bands. 73 Mike - KM0T www.km0t.com Comment John? Anyone else using Geralds mod kit with external reference? Results? ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz -- Laziness is the number one inspiration for ingenuity. Guilty as charged!
Re: [Flexradio] Frequency calibration etc
Mike King - KM0T wrote: I too have been using an external 10 MHz source on my SDR. First I used a Rubidium source off of ebay, never noticed anything different from the internal crystal other than no drift. (rock solid) Note that I have only used it on 902 thru 24 GHz - where the SDR is the 28 MHz IF. I recently switched the 10 MHz source to the HP Z3801 GPS receiver. I A/B switched the Rubidium and Z3801A back and forth and could tell no difference in the audio. However when I used the waterfall display while doing the switching, the Rubidium source show a bit of light white noise on the screen (where I adjusted it) as the waterfall was moving (it was very slight). The 10 MHz Z3801A source seemed to eliminate this. I am not experiencing anything drastic that would indicate the improper reception or weak signals covered up while using these 10 MHz sources, but I have no way to measure phase noise, just a gut feel for how my system is operating on the UHF/SHF bands. The small rubidiums aren't known for low phase noise (and some of them actually impose FM on the signal), so you might see an improvement when using the Z3801A which has a very low phase noise crystal. That might explain the difference you saw. Any electrically steered oscillator is subject to increased phase noise from noise on the steering signal, but the Z3801A seems to be pretty good. John
Re: [Flexradio] Frequency calibration etc
Well, I think we are a ways away from a kit. I was sort of suggesting that we try just using the Rockwell - Jupiter board and it's 10 mhz oscillator 'naked' and just see what we get. Give a look here : http://gpsdo.i2phd.com 73 Alberto I2PHD
Re: [Flexradio] Frequency calibration etc
And look here too: http://mysite.verizon.net/n1jez/osc/page5.html This board and Jupiter GPS module are available for $25, but apparently supplies are limited. A small number of us in the NJQRP club are doing a group project with it. You can see a close-up photo of a unit we had at the meeting this past weekend: http://www.njqrp.org/meetings/11-19-05/album/slides/img_2260.html . (You can see related photos by going to the photo index and selecting another thumbnail.) 73, George N2APB Well, I think we are a ways away from a kit. I was sort of suggesting that we try just using the Rockwell - Jupiter board and it's 10 mhz oscillator 'naked' and just see what we get. Give a look here : http://gpsdo.i2phd.com
Re: [Flexradio] Frequency calibration etc
Awesome Alberto! That is under an hours work to layout! We need to google and delve more deeply into our Italian genius' web site!! Bob Gollum wrote: Well, I think we are a ways away from a kit. I was sort of suggesting that we try just using the Rockwell - Jupiter board and it's 10 mhz oscillator 'naked' and just see what we get. Give a look here : http://gpsdo.i2phd.com 73 Alberto I2PHD ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz -- Laziness is the number one inspiration for ingenuity. Guilty as charged!
Re: [Flexradio] Frequency calibration etc
Robert McGwier wrote: Awesome Alberto! That is under an hours work to layout! We need to google and delve more deeply into our Italian genius' web site!! Bob, I am not the author of the project, it was developed by Andy Talbot G4JNT. I just hosted it in one of my pages. I have designed and built a GPSDO using a Motorola VP Oncore as GPS, an Isotemp OCX0-134-12 as oscillator and an Atmel AVR AT90S8535 as microcontroller. It works quite well (I have an HP Z3801A to compare it with), but I never had the time nor the will to fully document and publish the project on Internet. I do have a PDF file with the description of the project that I presented at a meeting on Digital Radio here in Italy, but unfortunately it is in Italian, so, unless you know this language, it is of little help. 73 Alberto I2PHD
Re: [Flexradio] Frequency calibration etc
John Thanks for the input. I have not heard of anyone else using the external reference, and you never noted any severe adverse effects. Is the 5065a keeping the SDR dead on, which is the primary objective in this whole discussion. Do you notice variation due to the Sound card clock? Perhaps all the theory we have been gumming about is not that important in practice. If you have to measure the difference on a high quality counter.. er well. Also thanks in advance for the tests. Thanks! -Original Message- From: John Ackermann N8UR [mailto:[EMAIL PROTECTED] Sent: Tuesday, November 22, 2005 8:35 AM To: ecellison Cc: 'Ross'; FlexRadio@flex-radio.biz Subject: Re: [Flexradio] Frequency calibration etc ecellison wrote: Ross Well, I think we are a ways away from a kit. I was sort of suggesting that we try just using the Rockwell - Jupiter board and it's 10 mhz oscillator 'naked' and just see what we get. In 2 lengthy arounds on this thread mostly what we have is theory and nothing tried at this point. I think that John - N8UR is the only one I have heard about who is using an external 10 mhz reference to the SDR. Comment John? Anyone else using Geralds mod kit with external reference? Results? Yes, I've been using my SDR-1000 with external reference from an HP 5065A Rubidium frequency standard. I haven't noticed any horrid phase noise problems from the 10MHz to 200MHz multiplication, but I also haven't been looking very hard for them. I hope that fairly soon -- perhaps over the Xmas holidays -- I will get temporary access to HP's latest super-duper phase noise measurement box, and one of the tests will be to compare the SDR-1000 with original crystal versus the external reference. I'll certainly post those results when I have them. 73, John
Re: [Flexradio] Frequency calibration etc
Alberto Yes, this is a really neat design. I think I have read it several times! Also a little embarrassed I have been calling the 10 khz Jupiter 10 MHZ. This is really pretty good short term stability for the GPS! I wonder if we could paste this into a Altera core? Eric -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Gollum Sent: Tuesday, November 22, 2005 10:25 AM To: FlexRadio@flex-radio.biz Subject: Re: [Flexradio] Frequency calibration etc Well, I think we are a ways away from a kit. I was sort of suggesting that we try just using the Rockwell - Jupiter board and it's 10 mhz oscillator 'naked' and just see what we get. Give a look here : http://gpsdo.i2phd.com 73 Alberto I2PHD ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
Re: [Flexradio] Frequency calibration etc
At 02:16 PM 11/22/2005, ecellison wrote: John Thanks for the input. I have not heard of anyone else using the external reference, and you never noted any severe adverse effects. Is the 5065a keeping the SDR dead on, which is the primary objective in this whole discussion. Do you notice variation due to the Sound card clock? Perhaps all the theory we have been gumming about is not that important in practice. On some on mobo sound cards I've tested, typical fractional variations in sample rate are around 2E-3 (that is, for a 48kHz sample rate, the actual sampling was at 48.090 kHz), and varied about 1E-3 over a time span of 2 seconds (that is, it went from 48.077 to 48.115 over that span). This is with a fairly high SNR, so one can really get 1Hz measurement uncertainty with a sample epoch of 100 milliseconds. Over that same 2 second time span, the deviation from a straight line trend was only around 20 ppm (1 sigma). On the other hand, some sound cards are much better: http://www.leapsecond.com/pages/sound-1pps/ Tom Van Baak reports a few tens of ppm accuracy. Several other pages report similar measurements (i.e. 100 ppm) I suspect that there is a wide variability in these things (mine were on mobo interfaces operating sitting out in the breeze, and probably a worst case) Jim, W6RMK
Re: [Flexradio] Frequency calibration etc
Well, you motivated me to do a simple test. I'm using the Baudline signal analysis program under Linux and clocking the sound card at (nominal) 96ksamples/second. I am currently injecting a 100mv p-p signal at 11.025kHz (to simulate the SDR-1000 IF output) into one channel of my Delta 44 card. The signal is coming from an HP 3325A synthesizer locked to an HP 5065A Rubidium standard. I'm using Baudline to decimate and downconvert the signal to get a few Hz spread across the screen (decimate by 2048, with an FFT size of 4096). Baudline has a high-precision frequency measurement function that has the potential of uHz resolution (it uses phase measurements rather than just FFT bins). Bottom line -- when I started the test half an hour ago, the nominal 11,025Hz signal showed as 11,024.33235 and right now, it shows as 11,024.33228. While that doesn't say anything about the long term stability, it's consistent with what I've observed over a year or more with this card; frequencies consistently read a little bit low, but never by more than a fraction of a Hertz. Now, this card is in a computer that's powered up 7x24 and has been running for months, so it's in an electrically and thermally stable environment. Results from a cold start might be different. 73, John ecellison said the following on 11/22/2005 05:16 PM: John Thanks for the input. I have not heard of anyone else using the external reference, and you never noted any severe adverse effects. Is the 5065a keeping the SDR dead on, which is the primary objective in this whole discussion. Do you notice variation due to the Sound card clock? Perhaps all the theory we have been gumming about is not that important in practice. If you have to measure the difference on a high quality counter.. er well. Also thanks in advance for the tests. Thanks! -Original Message- From: John Ackermann N8UR [mailto:[EMAIL PROTECTED] Sent: Tuesday, November 22, 2005 8:35 AM To: ecellison Cc: 'Ross'; FlexRadio@flex-radio.biz Subject: Re: [Flexradio] Frequency calibration etc ecellison wrote: Ross Well, I think we are a ways away from a kit. I was sort of suggesting that we try just using the Rockwell - Jupiter board and it's 10 mhz oscillator 'naked' and just see what we get. In 2 lengthy arounds on this thread mostly what we have is theory and nothing tried at this point. I think that John - N8UR is the only one I have heard about who is using an external 10 mhz reference to the SDR. Comment John? Anyone else using Geralds mod kit with external reference? Results? Yes, I've been using my SDR-1000 with external reference from an HP 5065A Rubidium frequency standard. I haven't noticed any horrid phase noise problems from the 10MHz to 200MHz multiplication, but I also haven't been looking very hard for them. I hope that fairly soon -- perhaps over the Xmas holidays -- I will get temporary access to HP's latest super-duper phase noise measurement box, and one of the tests will be to compare the SDR-1000 with original crystal versus the external reference. I'll certainly post those results when I have them. 73, John
Re: [Flexradio] Frequency calibration etc
At 03:31 PM 11/22/2005, John Ackermann N8UR wrote: Well, you motivated me to do a simple test. I'm using the Baudline signal analysis program under Linux and clocking the sound card at (nominal) 96ksamples/second. Bottom line -- when I started the test half an hour ago, the nominal 11,025Hz signal showed as 11,024.33235 and right now, it shows as 11,024.33228. While that doesn't say anything about the long term stability, it's consistent with what I've observed over a year or more with this card; frequencies consistently read a little bit low, but never by more than a fraction of a Hertz. Now, this card is in a computer that's powered up 7x24 and has been running for months, so it's in an electrically and thermally stable environment. Results from a cold start might be different. Excellent.. I suspect the Delta 44 has a halfway decent oscillator in it (unlike the mobo sound interface which almost certainly uses the processor clock or something divided down), and so, short term stability on the order of ppm is probably realistic, as you've measured. (0.0001 Hz out of 11kHz). Absolute accuracy wise, 0.7 Hz low out of 11kHz is, what, 58 ppm.. A typical inexpensive oscillator spec might be +/- 100 ppm as delivered... By the way, the baudline site does a bunch of analysis on some cards to show that they have certain good sample rates and others that are less good (because the oscillator doesn't evenly divide down to the required sample clock). Presumably this is not the case with the Delta 44 at 96 kHz. What's nice here is that once you've calibrated it, you're not going to have to fool around trying to track out short term variations (which is what I had to do for my particular application). 73, John Jim, W6RMK
Re: [Flexradio] Frequency calibration etc
You guys have probably already hashed this kind of thing out earlier, but does anyone at Flex have a relationship with the M-Audio people? Maybe we could use that channel to provide clock, sampling and phase noise improvement input directly to them... Of course if the D44 is only a short-term tactical decision this would be a moot point but if they were picked for more than just their currently acceptable card, maybe they'd listen. Mark -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Jim Lux Sent: Tuesday, November 22, 2005 5:57 PM To: ecellison; 'John Ackermann N8UR' Cc: FlexRadio@flex-radio.biz Subject: Re: [Flexradio] Frequency calibration etc At 02:16 PM 11/22/2005, ecellison wrote: John Thanks for the input. I have not heard of anyone else using the external reference, and you never noted any severe adverse effects. Is the 5065a keeping the SDR dead on, which is the primary objective in this whole discussion. Do you notice variation due to the Sound card clock? Perhaps all the theory we have been gumming about is not that important in practice. On some on mobo sound cards I've tested, typical fractional variations in sample rate are around 2E-3 (that is, for a 48kHz sample rate, the actual sampling was at 48.090 kHz), and varied about 1E-3 over a time span of 2 seconds (that is, it went from 48.077 to 48.115 over that span). This is with a fairly high SNR, so one can really get 1Hz measurement uncertainty with a sample epoch of 100 milliseconds. Over that same 2 second time span, the deviation from a straight line trend was only around 20 ppm (1 sigma). On the other hand, some sound cards are much better: http://www.leapsecond.com/pages/sound-1pps/ Tom Van Baak reports a few tens of ppm accuracy. Several other pages report similar measurements (i.e. 100 ppm) I suspect that there is a wide variability in these things (mine were on mobo interfaces operating sitting out in the breeze, and probably a worst case) Jim, W6RMK ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
Re: [Flexradio] Frequency calibration etc
I know that Gerald is a distributor since you don't sell M-Audio products without being one. So I suppose that constitutes a relationship. Bob Mark Amos wrote: You guys have probably already hashed this kind of thing out earlier, but does anyone at Flex have a relationship with the M-Audio people? Maybe we could use that channel to provide clock, sampling and phase noise improvement input directly to them... Of course if the D44 is only a short-term tactical decision this would be a moot point but if they were picked for more than just their currently acceptable card, maybe they'd listen. Mark -- Laziness is the number one inspiration for ingenuity. Guilty as charged!
Re: [Flexradio] Frequency calibration etc
Hi Mark, I have been day dreaming about a sound interface designed specifically for our needs. It could be a USB, Ethernet, or PCI device -- Ehternet being preferrable for its low cost, easy programming, and eletrical isolation. We could have a community developed interface with provisions for external clock or including a disciplined clock. We don't need any of the additional gadgets present in most sound interfaces. Just a simple high quality codec. One of the major advantages of such a device would be to acquire control over the repeatability and reproducibility of results. The Delta-44 browght us a large improvement. We could push the limits even further. 73, -- Edson Mark Amos wrote: You guys have probably already hashed this kind of thing out earlier, but does anyone at Flex have a relationship with the M-Audio people? Maybe we could use that channel to provide clock, sampling and phase noise improvement input directly to them... Of course if the D44 is only a short-term tactical decision this would be a moot point but if they were picked for more than just their currently acceptable card, maybe they'd listen. Mark -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Jim Lux Sent: Tuesday, November 22, 2005 5:57 PM To: ecellison; 'John Ackermann N8UR' Cc: FlexRadio@flex-radio.biz Subject: Re: [Flexradio] Frequency calibration etc At 02:16 PM 11/22/2005, ecellison wrote: John Thanks for the input. I have not heard of anyone else using the external reference, and you never noted any severe adverse effects. Is the 5065a keeping the SDR dead on, which is the primary objective in this whole discussion. Do you notice variation due to the Sound card clock? Perhaps all the theory we have been gumming about is not that important in practice. On some on mobo sound cards I've tested, typical fractional variations in sample rate are around 2E-3 (that is, for a 48kHz sample rate, the actual sampling was at 48.090 kHz), and varied about 1E-3 over a time span of 2 seconds (that is, it went from 48.077 to 48.115 over that span). This is with a fairly high SNR, so one can really get 1Hz measurement uncertainty with a sample epoch of 100 milliseconds. Over that same 2 second time span, the deviation from a straight line trend was only around 20 ppm (1 sigma). On the other hand, some sound cards are much better: http://www.leapsecond.com/pages/sound-1pps/ Tom Van Baak reports a few tens of ppm accuracy. Several other pages report similar measurements (i.e. 100 ppm) I suspect that there is a wide variability in these things (mine were on mobo interfaces operating sitting out in the breeze, and probably a worst case) Jim, W6RMK ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz ___ FlexRadio mailing list FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz
